torbjorn/my-pal-mcp-server
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: GPT-5.2 support

Browse Source
This commit is contained in:
Fahad
2025-12-11 19:11:50 +00:00
parent 5c3dd75ca6
commit 8b16405f06
33 changed files with 216 additions and 119 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
docs/custom_models.md
View File

@@ -61,7 +61,7 @@ The curated defaults in `conf/openrouter_models.json` include popular entries su
| `llama3` | `meta-llama/llama-3-70b` | Large open-weight text model |
| `deepseek-r1` | `deepseek/deepseek-r1-0528` | DeepSeek reasoning model |
| `perplexity` | `perplexity/llama-3-sonar-large-32k-online` | Search-augmented model |
| `gpt5.1`, `gpt-5.1`, `5.1` | `openai/gpt-5.1` | Flagship GPT-5.1 with reasoning and vision |
| `gpt5.2`, `gpt-5.2`, `5.2` | `openai/gpt-5.2` | Flagship GPT-5.2 with reasoning and vision |
| `gpt5.1-codex`, `codex-5.1` | `openai/gpt-5.1-codex` | Agentic coding specialization (Responses API) |
| `codex-mini`, `gpt5.1-codex-mini` | `openai/gpt-5.1-codex-mini` | Cost-efficient Codex variant with streaming |
@@ -77,21 +77,21 @@ View the baseline OpenRouter catalogue in [`conf/openrouter_models.json`](conf/o
Native catalogues (`conf/openai_models.json`, `conf/gemini_models.json`, `conf/xai_models.json`, `conf/dial_models.json`) follow the same schema. Updating those files lets you:
- Expose new aliases (e.g., map `enterprise-pro` to `gpt-5-pro`)
- Expose new aliases (e.g., map `enterprise-pro` to `gpt-5.2-pro`)
- Advertise support for JSON mode or vision if the upstream provider adds it
- Adjust token limits when providers increase context windows
### Latest OpenAI releases
OpenAI's November 13, 2025 drop introduced `gpt-5.1`, `gpt-5.1-codex`, and `gpt-5.1-codex-mini`, all of which now ship in `conf/openai_models.json`:
OpenAI's November 13, 2025 drop introduced `gpt-5.1-codex` and `gpt-5.1-codex-mini`, while the flagship base model is now `gpt-5.2`. All of these ship in `conf/openai_models.json`:
| Model | Highlights | Notes |
|-------|------------|-------|
| `gpt-5.1` | 400K context, 128K output, multimodal IO, configurable reasoning effort | Streaming enabled; use for balanced agent/coding flows |
| `gpt-5.2` | 400K context, 128K output, multimodal IO, configurable reasoning effort | Streaming enabled; use for balanced agent/coding flows |
| `gpt-5.1-codex` | Responses-only agentic coding version of GPT-5.1 | Streaming disabled; `use_openai_response_api=true`; `allow_code_generation=true` |
| `gpt-5.1-codex-mini` | Cost-efficient Codex variant | Streaming enabled, retains 400K context and code-generation flag |
These entries include pricing-friendly aliases (`gpt5.1`, `codex-5.1`, `codex-mini`) plus updated capability flags (`supports_extended_thinking`, `allow_code_generation`). Copy the manifest if you operate custom deployment names so downstream providers inherit the same metadata.
These entries include pricing-friendly aliases (`gpt5.2`, `codex-5.1`, `codex-mini`) plus updated capability flags (`supports_extended_thinking`, `allow_code_generation`). Copy the manifest if you operate custom deployment names so downstream providers inherit the same metadata.
Because providers load the manifests on import, you can tweak capabilities without touching Python. Restart the server after editing the JSON files so changes are picked up.